WGAN-TS

PyTorch implementation of A Two-Step Computation of the Exact GAN Wasserstein Distance, ICML 2018.

Usage

DATASET: mnist/cifar10/lsun
DATASET_PATH: path to the root folder of a dataset

1. Install dependency: CVXOPT for linear programming.
   
2. Download your dataset, unzip it and put it in DATASET_PATH.
   
3. run code using

python WGAN_TS.py --dataset DATASET --dataroot DATASET_PATH --BN_G

You are welcome to cite our work using:

@inproceedings{liu2018two,
  title={A Two-Step Computation of the Exact GAN Wasserstein Distance},
  author={Liu, Huidong and Xianfeng, GU and Samaras, Dimitris},
  booktitle={International Conference on Machine Learning},
  pages={3165--3174},
  year={2018}
}